Analysis of a random digraph generation technique used in the evaluation of parallel-processing static task allocation heuristics

One method of empirically evaluating static task allocation heuristics is to apply them to a sample of randomly generated task systems and compare the projected performance of the resulting allocations in a statistical manner with that of more established allocation heuristics. The validity of such comparisons is very much dependent on the manner in which the random task systems are generated. The authors analyze the p-method for random acyclic digraph generation empirically in terms of the expected sequentiality of randomly generated digraphs as the p parameter is varied, using mathematical entities called alignment functions and extensive computer simulation. This leads to a more complete understanding of the limitations (and possible misuses) of this methodology.